Re: kernel command line

[Roland McGrath]

> That's correct, but in the model of just "return null if the parent is
> gone", why is this a problem?

This model makes it impossible to guarantee an assignment of the task to
its original Hurdish ancestor.  I think that guarantee is the most valuable
part of the feature, because it allows unregistered Mach tasks to be
controlled by a non-root user who created them.